Improve test assertions

[jake-arkinstall]

At present, there exist tests that make np.isclose assertions on fidelity outcomes.

This means that an update that is beneficial to fidelity could result in failing tests. I have this problem on my machine, for instance.

I believe a good course of action is thus:

• Use one-sided bounds on assertions that provide metrics that we are happy to see improvements on
• Use a benchmarking approach alongside the one-sided bounds testing to track those metrics and report when there is an improvement or deterioration.

[jake-arkinstall]

Half done.

Now using upper bounds. It's clear we want benchmarks though so we can see when there's a performance degradation.